In a groundbreaking development, researchers at ETH Zurich have made a big leap in artificial intelligence, demonstrating that AI can now outperform humans in tasks requiring physical skills. This breakthrough was showcased through their AI robot, CyberRunner, which mastered the labyrinth marble game, a test of dexterity and precision, in a remarkably short time.

The labyrinth game, traditionally a test of human motor skills and spatial reasoning, involves guiding a marble through a maze-like board to succeed in a goal while avoiding pitfalls. This seemingly easy game demands considerable practice for humans to excel. Nevertheless, CyberRunner, developed at ETH Zurich and detailed on its dedicated website, achieved this feat in an unprecedented manner.

Using advanced model-based reinforcement learning, CyberRunner demonstrates how AI can extend its prowess into the realm of physical interaction. This system enables the AI to predict and plan actions by constantly learning from its environment. Equipped with a camera to look at the sport and motors to manage the board, the robot rapidly improved its gameplay through a process akin to human learning but at an accelerated pace.

Remarkably, CyberRunner accomplished its learning cycle in only over six hours, going through 1.2 million time steps at a control rate of 55 samples per second. This feat saw the AI surpass the record held by a highly expert human player by a powerful margin of over 6%.

Interestingly, during its learning phase, CyberRunner even discovered shortcuts in the sport, prompting the lead researchers, Thomas Bi and Prof. Raffaello D’Andrea, to intervene and guide the AI to avoid these paths.

This achievement by ETH Zurich researchers not only pushes the boundaries of AI in gaming but additionally signifies a serious step forward in how AI could be applied to real-world physical tasks. The success of CyberRunner indicates a future where AI can undertake complex physical activities, potentially transforming various industries and on a regular basis life.

This milestone in AI development marks a shift from virtual achievements, equivalent to mastering chess or Go, to conquering physical challenges, blurring the lines between human and machine capabilities within the realm of physical skill and dexterity.

A preprint of the research paper is accessible on the project website. As well as, Bi and D’Andrea will open source the project and make it available on the web site. Prof. Raffaello D’Andrea commented: “We imagine that that is the best testbed for research in real-world machine learning and AI. Prior to CyberRunner, only organizations with large budgets and custom-made experimental infrastructure could perform research on this area. Now, for lower than 200 dollars, anyone can engage in cutting-edge AI research. Moreover, once 1000’s of CyberRunners are out within the real-world, it would be possible to interact in large-scale experiments, where learning happens in parallel, on a worldwide scale. The final word in Citizen Science!”